{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:CH2BG5A2FQN6GR74Z3AE3IMZRL","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"ce59db6e76956b420d65cef75a44c38d490b7ed6d91ede8204652699cbfcb5fc","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-18T05:47:02Z","title_canon_sha256":"f6a5f1fa0958319e821d1d864492b8e2ee8d74fe924f148c61effc00405b4435"},"schema_version":"1.0","source":{"id":"2411.11296","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.11296","created_at":"2026-07-05T11:35:30Z"},{"alias_kind":"arxiv_version","alias_value":"2411.11296v2","created_at":"2026-07-05T11:35:30Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.11296","created_at":"2026-07-05T11:35:30Z"},{"alias_kind":"pith_short_12","alias_value":"CH2BG5A2FQN6","created_at":"2026-07-05T11:35:30Z"},{"alias_kind":"pith_short_16","alias_value":"CH2BG5A2FQN6GR74","created_at":"2026-07-05T11:35:30Z"},{"alias_kind":"pith_short_8","alias_value":"CH2BG5A2","created_at":"2026-07-05T11:35:30Z"}],"graph_snapshots":[{"event_id":"sha256:243f52e2ad6dc1a15c9b5c1277c5b08f3a7ead2d2649dc43603c21c17d554062","target":"graph","created_at":"2026-07-05T11:35:30Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2411.11296/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Responsible deployment of language models requires mechanisms for refusing unsafe prompts while preserving model performance. While most approaches modify model weights through additional training, we explore an alternative: steering model activations at inference time via amplifying sparse autoencoder (SAE) features that mediate refusal. This work uncovers a fundamental tension between SAE steering-based safety improvements and general model capabilities. While feature steering successfully improves robustness against both single-turn and challenging multi-turn jailbreak attempts, we discover","authors_text":"Blake Bullwinkel, David Majercak, Dean Carignan, Eric Horvitz, Forough Poursabzi-Sangdeh, Harsha Nori, Jingya Chen, Kyle O'Brien, Richard Edgar, Xavier Fernandes","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-18T05:47:02Z","title":"Steering Language Model Refusal with Sparse Autoencoders"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.11296","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:86d537baad865e21b9f42c5fe7f8a4679b6947412bda6f8d481f6d12d105286d","target":"record","created_at":"2026-07-05T11:35:30Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"ce59db6e76956b420d65cef75a44c38d490b7ed6d91ede8204652699cbfcb5fc","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-18T05:47:02Z","title_canon_sha256":"f6a5f1fa0958319e821d1d864492b8e2ee8d74fe924f148c61effc00405b4435"},"schema_version":"1.0","source":{"id":"2411.11296","kind":"arxiv","version":2}},"canonical_sha256":"11f413741a2c1be347fccec04da1998ac62b49c8d77c05a13fecee0843565416","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"11f413741a2c1be347fccec04da1998ac62b49c8d77c05a13fecee0843565416","first_computed_at":"2026-07-05T11:35:30.823505Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:35:30.823505Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ViuGKXRP7OLtzBZCD3JbTIFMrbCZ9tlzh6+tbWCnviCyrU0l3dItexgP9gdrUi/XD28zGl7u1fJ0z9l/TEhVAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:35:30.824044Z","signed_message":"canonical_sha256_bytes"},"source_id":"2411.11296","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:86d537baad865e21b9f42c5fe7f8a4679b6947412bda6f8d481f6d12d105286d","sha256:243f52e2ad6dc1a15c9b5c1277c5b08f3a7ead2d2649dc43603c21c17d554062"],"state_sha256":"573844aa09eb403d21f9ac2aa5f980e7d552251e6e5326e6691423a3de1cb09e"}